Reproducibility of peripapillary retinal nerve fiber thickness measurements with stratus OCT in glaucomatous eyes

Purpose: To determine the reproducibility of Stratus OCT peripapillary retinal nerve fiber layer (RNFL) measurements in glaucomatous eyes. Design: Experimental study. Participants: Fifty-one stable glaucoma patients with a range of severity. Methods: Peripapillary RNFL thickness was measured using the Standard and Fast scan protocols of Stratus optical coherence tomography (OCT) 3 times on the same day to determine intrasession variability and on 5 different days within a 2-month period to determine intersession variability. The same instrument was used by the same operator for all scans. Main Outcome Measures: Intraclass correlation coefficient (ICC), coefficient of variation (COV), and test-retest variability. Results: For mean RNFL thickness, the intrasession and intersession ICCs for the Standard and Fast scans were 0.98 and 0.96, respectively. The COV ranged from 3.8% to 5.2%. Test-retest variability was approximately 7 gm between sessions, most of which can be attributed to the approximate 5-mu m variability within each session. For quadrants, the ICC was 0.9 or higher and the COV was under 10% except nasally. Test-retest variability for quadrant measurements ranged from 6 to 16 mu m. For clock hours, test-retest variability approached 20 mu m between sessions in some sectors. In general, the ICC was lower in the nasal region than elsewhere. Variability was greater the smaller the area over which RNFL thickness was determined. Intrasession variability was not a predictor of intersession variability in individual subjects (P Standard = 0.72, P Fast =0.28). There was no relationship between variability and mean RNFL thickness (P Standard = 0.28, P Fast = 0.93). Conclusions: The reproducibility of Stratus OCT for RNFL thickness is sufficiently good to be useful clinically as a measure of glaucoma progression. When comparing 2 mean RNFL values on different days in the same eye, an 8-mu m decrease in thickness might be accepted as within normal limits of test-retest variability with 95% tolerance. For quadrants and clock-hour sectors, variability is higher, and more detailed calculations are necessary.